Mention rate
The share of tested questions whose answers mention your brand.
Methodology · v1.1
How we measure AI visibility.
We test buyer questions across ChatGPT, Gemini, Perplexity, Grok and Claude, then analyse brand mentions and cited sources. Here is how we choose the questions, record results and interpret the findings.
Definition
What "AI visibility" means.
AI visibility describes how often and how accurately a company appears in AI answers about its market.
Answers vary between runs and draw on multiple sources. We check a defined set of questions daily and compare results over time.
AEO and GEO
Answer engine optimisation (AEO) and generative engine optimisation (GEO) describe work to improve a brand's presence in AI answers.
The question list
How we build the list of questions.
We build a list of up to 75 questions from conversations with your sales team. Questions are grouped into three segments by topic or audience and cover four stages of buying:
- 01
Problem-aware
The buyer is describing a problem. For example, "Why is our reporting always out of date?"
- 02
Solution-aware
The buyer is exploring a type of solution. For example, "What does a data warehouse project involve?"
- 03
Comparing suppliers
The buyer is choosing between suppliers. For example, "Best analytics consultancies for mid-market SaaS".
- 04
Asking about you by name
The buyer asks about your brand or compares it directly with a competitor.
You review and approve the questions before tracking starts.
Question coverage
The standard service tracks 75 active questions daily. Questions can be replaced as priorities change. Multi-market services support up to 1,000 questions for additional products, audiences or regions.
Engines
AI assistants covered
ChatGPT, Gemini, Perplexity, Grok and Claude.
We keep the same engines within each reporting period. Changes are recorded in the log below and in the weekly report.
What we record
What we record on every run.
Citation share
The share of cited sources that are pages on your own website.
Share of voice
How often your brand appears compared with your chosen competitors.
Tone
How the answers describe your brand, including praise, criticism or warnings.
Position
Where your brand appears within an answer.
Source websites
The websites cited in each answer.
Crawler traffic
Which AI crawlers visit your website and how often.
We check questions daily and report weekly trends to reduce the effect of individual variations.
Sources
How we review cited sources
AI assistants may cite review sites, directories, publications, comparison pages, forums and company websites.
We list the sources cited in answers about your category and check where your brand appears. This helps identify listings to correct, missing coverage and content to improve.
Technical checks
Two checks on your own website.
Can AI tools read your site?
We check whether crawler rules or security settings block AI tools from reaching your website. Findings identify the affected pages and settings.
Do your company details match?
We compare your company name, description, founders and location across your website and cited sources, and report discrepancies that need correction.
The limits of this method.
Question coverage
We test a selected set of questions. The results do not cover every question a buyer could ask.
Variation between answers
The same question can return different answers. Repeated checks help identify patterns; individual results may vary.
Region and personalisation
Answers can vary by country, language and user history. We control the region within an engagement but cannot reproduce every user's experience.
Revenue attribution
A change in visibility does not establish its effect on sales. We state the assumptions behind attribution and do not attempt it in the first six months.
Model updates
A shift across a category may suggest a model update. We label this as an inference unless the provider confirms it.
No guaranteed mentions
AI providers control their models. Our findings and recommended fixes cannot guarantee what an assistant will say.
Change log
Method change log.
| Date | Change |
|---|---|
| September 2026 · v1.1 | Daily checks of 75 active questions across ChatGPT, Gemini, Perplexity, Grok and Claude, with weekly reporting. |
| September 2026 · v1.0 | First published version of the method. |
See the method run on your own brand.
The Diagnostic applies everything on this page to your company, your competitors and your buyers' questions.